Archived

This topic is now archived and is closed to further replies.

dbrooks

Memory timings @ 4gHz

Recommended Posts

Hello All,Anybody able to get 7-7-7 timings @1600 with DDR3 @ 4gHz on a P6T deluxe with an i7920? I can't seem to get thatto be stable on my board. I can post voltages if needed. See sig for specs. Thanks,Dan

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

While some OC on this forum, you are vastly wrong here, as the chance to get the right answer are pretty slim.Suggest you go here and post your question:http://www.xtremesystems.org/forums/index.phpBut, my opinion, boards have limits. And this might be one. For example, if I clock my board over 400mhz, I get stable system and CPU, but my winrar is giving me troubles. Go figure.Good luck!

Share this post


Link to post
Share on other sites
While some OC on this forum, you are vastly wrong here, as the chance to get the right answer are pretty slim.Suggest you go here and post your question:http://www.xtremesystems.org/forums/index.phpBut, my opinion, boards have limits. And this might be one. For example, if I clock my board over 400mhz, I get stable system and CPU, but my winrar is giving me troubles. Go figure.Good luck!
Thanks Word Not Allowed, I'll check around. Have thought about trying 1800 or 2000 mhz memory when the prices start to come down a bit too. I realize it's not that much of a gain, but I do seem to see a difference at the faster timings at that clock. Just doesn't last too long before I get problems. It may just be the board too as you say.

Share this post


Link to post
Share on other sites
Hello All,Anybody able to get 7-7-7 timings @1600 with DDR3 @ 4gHz on a P6T deluxe with an i7920? I can't seem to get thatto be stable on my board. I can post voltages if needed. See sig for specs. Thanks,Dan
Is the memory rated for 7-7-7?If so the deal is the 920 itself and it will require raising QPI/DRAM to stabilize it. Assumung you DRAM voltage is 1.65-1.67, set QPI/DRAM to 1.40 and try again. But do not go any higher than 1.40 on QPI without discussing it with me first. The issue is the 920 as they have discovered this chip is showing different results for stability in clocking due to memory.. the 940 and 965 do not display the problem anywhere near as much. The the QPI/920 problem was first discovered and discussed by eva2000. It may not be possible to stabilize the timing @ above 3.8Ghz with the 920.. it really depends on the processor. Intel is releasing an updated version of all their procs with better clocking stability ... I do believe the D0 stepping replacement for the 920 is already out as of March 2nd. the 940 and 965 will be replaced by the 950 and 975
While some OC on this forum, you are vastly wrong here, as the chance to get the right answer are pretty slim.Suggest you go here and post your question:http://www.xtremesystems.org/forums/index.phpBut, my opinion, boards have limits. And this might be one. For example, if I clock my board over 400mhz, I get stable system and CPU, but my winrar is giving me troubles. Go figure.Good luck!
I happen to be one of 'those guys' over there who works with the manufactures on testingbut I do not go by the same screen name His board is fine. its the 1st gen 920 which may or may not be fixable with QPI voltage

Share this post


Link to post
Share on other sites
Is the memory rated for 7-7-7?If so the deal is the 920 itself and it will require raising QPI/DRAM to stabilize it. Assumung you DRAM voltage is 1.65-1.67, set QPI/DRAM to 1.40 and try again. But do not go any higher than 1.40 on QPI without discussing it with me first. The issue is the 920 as they have discovered this chip is showing different results for stability in clocking due to memory.. the 940 and 965 do not display the problem anywhere near as much. The the QPI/920 problem was first discovered and discussed by eva2000. It may not be possible to stabilize the timing @ above 3.8Ghz with the 920.. it really depends on the processor. Intel is releasing an updated version of all their procs with better clocking stability ... I do believe the D0 stepping replacement for the 920 is already out as of March 2nd. the 940 and 965 will be replaced by the 950 and 975I happen to be one of 'those guys' over there who works with the manufactures on testingbut I do not go by the same screen name His board is fine. its the 1st gen 920 which may or may not be fixable with QPI voltage
Thanks for the info Nick. I did not know that about the 920's stability issue. I have not tried as high as 1.40 on the QPI.The memory is the OCZ 1600 Platinum with 7-7-7-24 timings. http://www.newegg.com/Product/Product.aspx...=OCZ3P1600LV6GKI'll give it a go as soon as my GTX285 step-up arrives sometime next week.Thanks again,Dan

Share this post


Link to post
Share on other sites
Thanks for the info Nick. I did not know that about the 920's stability issue. I have not tried as high as 1.40 on the QPI.The memory is the OCZ 1600 Platinum with 7-7-7-24 timings. http://www.newegg.com/Product/Product.aspx...=OCZ3P1600LV6GKI'll give it a go as soon as my GTX285 step-up arrives sometime next week.Thanks again,Dan
Verify your DRAM voltage first .. it can go to 1.67 with no issues at all.. and I may give you another BIOS setting to try if QPI/DRAM 1.40v does not helpIf it does.. then your job will be to back the QPI/DRAM down (1.39, 1.38, 1.37) intil it is unstable and lock it at the voltage it is stable at. 1.40 is OK but do not go any higher without posting and discussing first as I will probably add another setting to try.

Share this post


Link to post
Share on other sites
Verify your DRAM voltage first .. it can go to 1.67 with no issues at all.. and I may give you another BIOS setting to try if QPI/DRAM 1.40v does not helpIf it does.. then your job will be to back the QPI/DRAM down (1.39, 1.38, 1.37) intil it is unstable and lock it at the voltage it is stable at. 1.40 is OK but do not go any higher without posting and discussing first as I will probably add another setting to try.
Ok, will give it a try. Having trouble though verifying my DRAM voltage. I can set it in the BIOS at say 1.64, 1.66 or even 1.68 etc, butusing HWMonitor shows 1.66 almost no matter what I set the voltage in the BIOS. Is there a good way to verify the DRAM voltage?TurboV, the ASUS utility, shows whatever I have set in the BIOS for voltages.Will be probably be toward the end of this week before I start messing with it again as I will prefer to have the final hardware setup withthe new graphics card installed.Don't seem to have a problem really at 3.8 gHz and 7-7-7, though I've not run a long Prime test at this clock, maybe only 4 hours or so.Currently back on the 'ol P4 and boy, I have really missed my new system.Thanks again for your help in this matter. I feel like I'm so close to getting the thing stable at 7-7-7 timings, but there's something just not quite right.

Share this post


Link to post
Share on other sites
Don't seem to have a problem really at 3.8 gHz and 7-7-7
Thats how I know this is the 920 issue which eva2000 stumbled onAnything above 3.8 is iffy getting memory stable without making concessions for timing. Some have a hard time getting above 3.8 at all.I do not trust those HW monitors. Set it in the BIOS directly DRAM to 1.66 to 1.68 (is fine) and QPI/DRAM 1.36 to 1.42 is fineAs I recall eva2000 had to run a 1.42v QPI on one DFI board to stabilize memory a 4G clockI do not use 920's ... I knew some time ago they were prone to being hot and finicky and sometimes difficult above 3.8 which is why I suggest the 940 instead. Its the luck of the draw on getting a 920 that is easy to clock with no issues above 3.8There is a spec which Intel released which is important.. and that is QPI/DRAM (UNCORE) must be 'within' .5v of DRAMIn other words, if QPI/DRAM is too LOW and is less than .5v of DRAM processor damage can occur. So the minimum QPI/DRAM voltage for a 1.66 DRAM setting is 1.16v. As long as QPI/DRAM remains within that .5v range of DRAM its safe. Right now memory companies are running systems at 1.5-1.65v QPI/DRAM and 1.65v DRAM for DDR3 1866-2000 memory and to be quite honest there are a lot of people questioning that.. including myself LOL So I wish to be cautious about making voltage suggestions with this clocking issue. I know 1.40 is not a problem at all and 1.43 would also more than likely be OK but once you start pushing higher the long term real-world result is really unknown.We are also finding there may be a relation between the QPI/DRAM and Vcore as well with higher memory speeds (1800 and above) however that will not apply in this case. Do make sure you have correctly set up the BIOS in shutting down the advanced CPU menu items Intel SpeedStep Tech DisabledIntel Turbo Mode Tech DisabledCPU Spread Spectrum DisabledPCI Spread Spectrum DisabledC1E Support DisabledIntel Virtualization Tech DisabledCPU TM Function DisabledIntel HT Technology Disabled <--- Hyperthreading (Note: Enabled for video encoding., disabled for coolest temp in clock as FSX has no i7 type HT support)Intel C-State Technology: Disable And I assume you are manually setting the memory timing in the BIOS by the listDRAM CAS# Latency 7DRAM RAS# to CAS# Delay 7DRAM RAS# PRE Time 7DRAM RAS# ACT Time 20

Share this post


Link to post
Share on other sites

OK, got the system prime95 stable for 6hrs 38mins before I stopped it manually. Had to use 1.36vcore, 1.40v QPI, and 1.68vDRAM. Still have to test at lower QPI settings, i. e. 1.38 etc, but if the DRAM is set anywhere below 1.68v, i get errors almost immediately in Prime. Same thing with the vcore, 1.35v just doesn't cut it. Core temps topped out at 70C running mostly about 68C. The 1.40QPI setting and a bit more on the vcore and DRAM voltages definitely make the difference for 7-7-7-24 @ 1600 and 4gHz. Mind you, this will not be a 24/7 clock, but will try for FSX usage. I can run 8-8-8-24 timings 24/7 @ 4ghz and 1600 no problem at all with lower voltages all around.Thanks again Nick, you have taught me something more about my system and system tweaking.EDIT: I bought 7-7-7 memory and that's what I wanted to get out of it. I have, so I'm happy about that. As far as FSXgoes, I do seem to see a difference at these timings over the 8-8-8 timings. How much of this is "subjective enthusiasm" I can't say. But there seems to be a noticeable difference (smoothness). I'm not smart enough to quantify the results.

Share this post


Link to post
Share on other sites
OK, got the system prime95 stable for 6hrs 38mins before I stopped it manually. Had to use 1.36vcore, 1.40v QPI, and 1.68vDRAM. Still have to test at lower QPI settings, i. e. 1.38 etc, but if the DRAM is set anywhere below 1.68v, i get errors almost immediately in Prime. Same thing with the vcore, 1.35v just doesn't cut it. Core temps topped out at 70C running mostly about 68C. The 1.40QPI setting and a bit more on the vcore and DRAM voltages definitely make the difference for 7-7-7-24 @ 1600 and 4gHz. Mind you, this will not be a 24/7 clock, but will try for FSX usage. I can run 8-8-8-24 timings 24/7 @ 4ghz and 1600 no problem at all with lower voltages all around.Thanks again Nick, you have taught me something more about my system and system tweaking.EDIT: I bought 7-7-7 memory and that's what I wanted to get out of it. I have, so I'm happy about that. As far as FSXgoes, I do seem to see a difference at these timings over the 8-8-8 timings. How much of this is "subjective enthusiasm" I can't say. But there seems to be a noticeable difference (smoothness). I'm not smart enough to quantify the results.
OK, well here is a little tid bit.. but dont do this!.. the max on DRAM 1.85v not 1.65 ... the spec Intel posted for absolute max memory voltage has been revised however its still being tested to verify. Right now we know for a fact that 1.75 is safe and 1.80-1.85 is being loop run in tests. We know above 1.85 is burning out procs for sure. None the less keep it under 1.70 for right now till all tests are concluded and the results get posted at places like AnandtechAs for QPI/DRAM.. once you verify stability start bringing it down till it hits 1.35 and stop. It is possible your board is simply out of calibration on the DRAM voltage and needed a touch more to compensate. Regardless, the newer i7 are running DDR3 2000+ on 1.37 QPI so the QPI voltage issue is most likely tied to the 1st generation i7 processors.

Share this post


Link to post
Share on other sites
OK, well here is a little tid bit.. but dont do this!.. the max on DRAM 1.85v not 1.65 ... the spec Intel posted for absolute max memory voltage has been revised however its still being tested to verify. Right now we know for a fact that 1.75 is safe and 1.80-1.85 is being loop run in tests. We know above 1.85 is burning out procs for sure. None the less keep it under 1.70 for right now till all tests are concluded and the results get posted at places like AnandtechAs for QPI/DRAM.. once you verify stability start bringing it down till it hits 1.35 and stop. It is possible your board is simply out of calibration on the DRAM voltage and needed a touch more to compensate. Regardless, the newer i7 are running DDR3 2000+ on 1.37 QPI so the QPI voltage issue is most likely tied to the 1st generation i7 processors.
Not that I plan to attempt this until the green light has been given...but will raising the DRAM to 1.75v do anything to allow higher clocks or lower timings to an existing set of dominator DDR3 1600s?-jk

Share this post


Link to post
Share on other sites
Not that I plan to attempt this until the green light has been given...but will raising the DRAM to 1.75v do anything to allow higher clocks or lower timings to an existing set of dominator DDR3 1600s?-jk
It will but that is not the primary issue with 1st gen procs above 1600.. its QPI/DRAM (AKA UNCORE or CPU VTT on some motherboards)In the case of the older 1600 memory chips running 1866 (not the new ones on the market) it requires a min CAS8 8-8 with correct subs, a QPI/DRAM of 1.50-1.60 and possibly some vDRAM. Even the newer memory products running CAS7 are the same with voltage and 1st gen i7none the less, that is well outside of Intel spec on QPI which has a max of 1.37 but the memory companies are specifying that for their 1866 products and 1st gen i7 which I highly question regardless of what the OC nuts are doing and saying its safeMy intuition tells me they are designing memory with the new chips (1866-2000+) for the 2nd gen batch of i7 and anyone messin with a 1st gen is SOL if it burns because its considered a clock outside of spec

Share this post


Link to post
Share on other sites